| I think some other PCI cards are able to achieve this (or very close to) latency as well.
Some FW interfaces (MoTU was one that we measured) have very good latency characteristics. The 896 reports 4 samples of latency (each on in and out) and 24 samples of safety offset (each on in and out). So, for a 16 sample I/O buffer, that is 4 + 24 + 16(input) + 4 + 24 + 16(output) = 88 samples, or about 2 msec in to out when run at 44.1kHz. We've measured this at 32 sample I/O in the past (which was 2.7msec), but I don't see any reason it couldn't be run at 16 sample I/O if you are able to do the work you need at this granularity.
Bill On Jul 6, 2012, at 4:41 PM, Markus Fritze wrote: With CoreAudio and the Build-In audio hardware @ 44100Hz with 16 samples buffer size we did measure a round-trip latency of about 2ms. Measured, not guessed. The only other hardware we found that is able to get 16 samples under CPU load is the Apogee Symphony, but you won't get better than the internal audio hardware.
Markus On Jul 6, 2012, at 3:36 PM, Paul Davis < email@hidden> wrote:
On Fri, Jul 6, 2012 at 4:19 PM, David Reaves <email@hidden> wrote:
For some time now, Windows using ASIO has provided users with millisecond-range latency. If by "dedicated hardware" you include a huge number of run-of-the mill PCI audio cards, then you are correct.
this is only true if by "millisecond range" you mean "several milliseconds" and if by latency you mean "playback latency" and by "provided users" you mean "provided users whose non-audio h/wand other software doesn't get in the way".
there's a large amount of varyingly-inaccurate, misleading info that circulates around regarding various kinds of latency on various general purpose OS's. i think its very important to be clear about all the caveats and details.
|